My WIPs are branching off

[joelreymont]

How can I straighten this out into a linear sequence with no branches?

@  kpxxpoyk [email protected] 2025-01-02 18:56:43 master* 9377fda1
│  wip
│ ◆  kpxxpoyk hidden [email protected] 2025-01-02 18:00:24 master@origin 40e56931
├─╯  wip
◆  mxmtyrwl [email protected] 2025-01-02 17:27:11 863571c4
│  Parse context actions

Also, is jj meant to show only the last couple of commits?

[jennings]

Notice that kpxxpoyk is marked “hidden”. That indicates that you’ve rewritten that commit since it was pushed, from 40e56931 to 9377fda1.

The usual way to linearize this is to rebase onto master@origin, but that will create a divergent change because it will unhide 40e56931. This isn’t bad, it just needs to be dealt with.

There are two ways you might linearize this. You could duplicate the second revision to get a new change ID, then rebase:

jj duplicate 9377fda1
jj edit NEW_CHANGE_ID
jj abandon 9377fda1
jj rebase -d master@origin

Or you could restore the contents of 9377fda1 into a new change:

jj new master@origin
jj restore --from 9377fda1
jj abandon 9377fda1

Regarding your second question, see this FAQ, you can configure revsets.log to change ehat is logged by default.

[joelreymont]

I didn't manage to execute the above well enough and ended up hard-resetting the repo using git. Then I got myself into the same situation, again.

I fully understand what's going on now but still don't understand how to fix this using jj.

@  wlrqtmpm [email protected] 2025-01-06 18:40:56 master* cf525331
│  Parse lvalue
│ ◆  wlrqtmpm hidden [email protected] 2025-01-06 17:44:27 master@origin 15771b35
├─╯  wip
◆  pvqvtnxv [email protected] 2025-01-02 17:27:11 863571c4
│  Parse context actions

The above happens because I work on a bunch of changes and push them to Github as work in progress. Then I work some more and have a batch of changes that supersedes the previous one. I would just amend the commit and force-push it to Github using git. How do I do this using jj, though?

The original instructions ask me to abandon the latest change if I understand them correctly. I want to keep that change, though, and drop the previous one. That is, I'd like to keep 9377fda1 and drop 40e56931 instead. How do I do this?

Abandoning 15771b35 doesn't seem to work.

❯ jj abandon 15771b35 --ignore-immutable
Abandoned commit wlrqtmpm hidden 15771b35 master@origin | wip
Nothing changed.
❯jj log
@  wlrqtmpm [email protected] 2025-01-06 18:40:56 master* cf525331
│  Parse lvalue
│ ◆  wlrqtmpm hidden [email protected] 2025-01-06 17:44:27 master@origin 15771b35
├─╯  wip
◆  pvqvtnxv [email protected] 2025-01-02 17:27:11 863571c4
│  Parse context actions

[martinvonz]

What's your revsets.log set to? We don't include remote_bookmarks() by default but it looks like you've added that to your version and that's why you see hidden commits in jj log. Maybe that's what confused you? Maybe there's a better default revset for you.

[scott2000]

I've had this happen actually because trunk() is included by default, so if you do --ignore-immutable and rewrite trunk, the hidden commit still shows up.

[joelreymont]

I'm using the default settings and haven't added anything.

What confused me is seemingly not being able to abandon the old commit that I have pushed and then edited. Running jj git push, after abandoning, does remove it, though.

[martinvonz]

Oh, interesting. It's a pretty unusual case then. It's probably fine to include trunk() even when it's hidden then.

[jennings]

@joelreymont Ah, sorry, my instructions weren't quite correct because I was typing them from my phone. I thought jj duplicate also began editing the new commit, but it doesn't.

[joelreymont]

@martinvonz It would be helpful if the abandoned commit disappeared before jj git push, otherwise it's confusing. You abandon a commit but it stays put 🤷🏼‍♂️.

[scott2000]

Could be a bad habit but this is my personal and private repo so I use force-push to back up my work. I suppose I could squash a bunch of commits together but force-push seems simpler.

I do this too actually! For personal repos, I like to set immutable_heads() to none() to allow mutating any commit without using --ignore-immutable, and I also change the log revset to something else like ancestors(trunk()----.. | present(@), 2) to only show 5 commits from trunk() generally. I'm sure there's lots of other configurations that work for this, but that's what I use personally since I rewrite history a lot in private repos.

[jennings]

Understood! I wasn't scolding, the only reason I asked was because of what @scott2000 said: You probably want to update your immutable_heads() definition in this repo so jj will let you rewrite commits on the master branch without needing --ignore-immutable.

[joelreymont]

If simply doing jj git push as suggested in this comment did what you want, then you don't need to abandon anything.

This works!

[joelreymont]

How do you check the repo config into the repo?

Does this happen automatically?

[bnjmnt4n]

Repo-specfic config isn't currently checked into the repo itself. You'd have to copy the configuration to new machines. #3684 has a discussion about this.